This invention relates to systems and methods for connecting integrated circuits to circuit boards in a flexible and dense manner.
In a variety of situations, economic or functional considerations make it desirable to alter the set of integrated circuits used on a circuit board, such as in a personal or other computer. For example, it may become useful to add a new integrated circuit or to replace or alter the packaging of an existing integrated circuit. However, such changes are often difficult to achieve, because computer circuit boards are typically pre-designed for a particular set of integrated circuits. Thus, altering the set of integrated circuits often requires a board redesign, which can result in significant added costs, delays, or compatibility problems. Consequently, manufacturers often forego changes to the chip set, even if they would result in improved performance or reduce costs.
One particular instance in which it is desirable to use a different chip set but in which the difficulties of redesigning the board stand as obstacles, is the decision to alter the packaging of prepackaged, mass-produced integrated circuits. During the lifetime of such a chip, there may be several phases of packaging, and the best choice may also change over time, as detailed below.
Mass-market prepackaged semiconductor chips often use pin grid array ("PGA") packages, which have metal leads extending from the bottom surface of the package as rows of pins. The array of pins in PGA packages permits a large number of leads, which is essential for complex circuits with high input-output requirements. PGA packages typically allow more leads than alternatives such as dual in-line packages (which have two rows of pins along the edge of the package). A variety of PGA packages are described in my U.S. Pat. No. 4,750,092, which is hereby incorporated by reference.
High-volume, complex, prepackaged semiconductor chips, such as microprocessors, are typically introduced in a ceramic PGA package, even though that package type is relatively costly. Although ceramics have certain other advantages, such as good hermetic sealing, chip makers are principally willing to suffer the cost penalty of ceramics in early stages of chip sales because the packaged chip can be assembled onto circuit boards by virtually all board designers.
As chip volumes increase and chip manufacturing costs drop, chip makers generally look for ways to reduce the package cost. Often, they will begin offering a packaging option of plastic surface-mounted geometry, such as the plastic leaded chip carrier ("PLCC") or the plastic quad flat package ("PQFP"). Those surface-mounted packaging options can reduce the cost of the package by perhaps 30-70%, which can result in a packaged circuit, such as a microprocessor, that sells for 10-50% less than the ceramic-PGA-packaged version.
Unfortunately, not all chip purchasers have the capability and desire to take advantage of the lower-priced, plastic, surface-mounted packages. Some lack the surface-mounting equipment or technical knowledge to use a surface-mounted package. Others cannot or would not find it cost-effective to redesign their circuit board to accommodate the different "footprint" required by a surface-mounted package. Still others are discouraged by the disadvantages of surface-mounting using surface-mount packages, such as the cost of a fine-pitch socket (to which the surface-mounted package might need to be attached), the complications that can arise in board assembly when using surface-mounting for some circuits (as opposed to the ease of through-hole mounting of PGA packages), or the difficulty in testing installed surface-mounted circuits and replacing faulty ones.
No previously existing devices or methods exist to install the lower-cost surface-mounted chips on a circuit board without altering the board layout or increasing the difficulty of the installation and testing process from that common to the use of PGAs. Previous packaging techniques, therefore, have forced users incapable of surface mounting themselves to pay for the higher-cost PGA-packaged semiconductors.
Another problem with previous packaging techniques has been the difficulty in reliably assembling the package and affixing it to the circuit board, and with inspecting the connections. It would be desirable to be able to inspect the assembled device separately, before the circuit board is completed. Those goals are difficult with surface-mount techniques, because the packaged chip must be soldered directly to the circuit board, which can be difficult both to accomplish and to inspect electrically until the board is completed. PGA packages are easier to install because they simply plug in the board, and they can be tested before installation by plugging them into a test board. It has been difficult heretofore to make and inspect, however, the contact between each pin and the conductive pattern bonded out from the semiconductor chip.
Another instance in which it is desirable to use a different chip set but in which the difficulties of redesigning the board stand as obstacles, is the decision to add a coprocessor circuit, or another circuit that can be used with an integrated circuit such as additional memory, to the existing chip set. The optimal choice may change across a customer base or over time, as described below.
It has been common to place the circuitry to perform certain functions in a computer's CPU--typically functions that are not "core" ones or that may be needed at varying levels of performance--on separate chips from the microprocessor. Such chip, called "coprocessors," have commonly included mathematical coprocessors or graphical coprocessors. Often, the end user is permitted to purchase the coprocessor separately, if accelerated performance or added functions are needed. It is not uncommon for some users to select the coprocessor upon purchasing the computer, for others to purchase and have it installed after purchase, and for yet others to skip it entirely.
Such flexibility results in increased cost and trouble for the board designer, though. If any subset of customers will demand the coprocessor, the board designer must include a board location (typically a socket) for it in the design of the circuit board. Designing in a socket is costly in board space, particularly for a feature that only certain customers, perhaps even a minority, will use. If some customers will demand the coprocessor later on, it is not even possible to use two different board designs, one with the coprocessor and the other without, but even if that option were possible, it is costly. No current packaging or design techniques permit a single design for both types of systems, those with and those without coprocessors.
Another example of an instance in which it is desirable but difficult to alter the chip set, is the desirability of providing sockets or circuit board locations that are suitable for any of an interchangeable set of components. For example, some boards can be upgraded to a higher-powered microprocessor in a series merely by unplugging the old one and replacing it with a new one. Previous techniques for achieving such compatibility, however, have depended on the use by the chip maker of a common "pinout," where the pins on each of the interchangeable chips perform the same function. A truly flexible interchangeable system would permit compatibility with a wide variety of components, even ones with different pinouts.
Yet another example of an instance in which it is desirable but difficult to alter the chip set, is when an integrated circuit is replaced with an equivalent component drawing lower power. Lower-power upgrades are desirable because they create less heat, are less costly, and can run faster. However, to date, switching to a lower-powered circuit has required complete redesign of the circuit board, because associated components must be redesigned to work with the lower power, too, and because the changed circuit will not work in the originally designed board. It would be desirable to create a system that permitted lower-powered upgrades of an integrated circuit to be plug-compatible.
In addition, the invention has application in general circuit board design, in which the problem is not increased flexibility, but increased circuit density. Previous attempts to increase circuit board density by creating circuit board modules supporting chip sets or multi-level circuit boards have failed to produce reasonable-cost results. Common problems have included the failure of such techniques to use standardized components or the need for special materials or fabrication techniques. Furthermore, none of the previous packaging methods including PGA and surface-mount technology, have addressed the need to pack chips on the circuit board closer together without giving up the cost advantages of prepackaged, generally available, or standard chip components. Improved packing techniques can also lead to certain collateral benefits, including improving testing and manufacturability of the board assembly.